EUVIMEDEUVIMED entdecken
Crossref

Deep-Learning Based Deconvolution for Correcting Spatial Impulse Response of Transducer in Optoacoustic Tomography

Isha Munjal, Jaya Prakash

IEEE Transactions on Instrumentation and Measurement · 2024 · 73 · 1-12

DOI: 10.1109/tim.2024.3417542 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2024. Verfasst von Isha Munjal und 1 weiteren Autor:innen. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 59 Literaturangaben. Gefördert von: Indian Institute of Science Startup, Department of Biotechnology (DBT)—Innovative Young Biotechnologist Award (IYBA) Program, Science and Engineering Research Board (SERB)—Startup Research Grant.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The image quality in optoacoustic tomography (OAT) is affected by various physical parameters like geometry of the transducer, transducer response, bandwidth of the transducer, laser pulsewidth, and detector position. Due to the transducer response, the optoacoustic (OA) signals are typically distorted and the estimation of pressure distribution becomes inaccurate, which is apparent in the reconstructed OAT images. This work aims to improve the degraded OAT images by eliminating the effect of spatial impulse response (SIR) of the transducer array. This work modeled and corrected the SIR for two transducer array configurations namely focused-concave array and unfocused-linear array transducer. The comparison revealed that SIR affect was more prominent with the focused-concave array case compared to unfocused-linear array case. The SIR correction was performed with sparse recovery-based deconvolution scheme and deep-learning based deconvolution. The improvement in the OAT image quality was evaluated with structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR). The fully dense-UNet (FD-UNet) architecture was able to remove the SIR contribution from the OA images. Specifically, FD-UNet was able to improve SSIM by about 50%–80% and more than 10 dB in terms of PSNR compared to backprojection and sparse recovery-based deconvolution with in-silico data. Further FD-UNet improved localization and contrast (by about 25%) with experimental data.

Förderung

  • Indian Institute of Science Startup
  • Department of Biotechnology (DBT)—Innovative Young Biotechnologist Award (IYBA) Program
  • Science and Engineering Research Board (SERB)—Startup Research Grant

Zitieren

APA

Munjal, I., & Prakash, J. (2024). Deep-Learning Based Deconvolution for Correcting Spatial Impulse Response of Transducer in Optoacoustic Tomography. IEEE Transactions on Instrumentation and Measurement, 73, 1-12. https://doi.org/10.1109/tim.2024.3417542

Vancouver

Munjal I, Prakash J. Deep-Learning Based Deconvolution for Correcting Spatial Impulse Response of Transducer in Optoacoustic Tomography. IEEE Transactions on Instrumentation and Measurement. 2024;73:1-12. doi:10.1109/tim.2024.3417542

BibTeX
@article{Munjal2024,
  author = {Munjal, Isha and Prakash, Jaya},
  title = {Deep-Learning Based Deconvolution for Correcting Spatial Impulse Response of Transducer in Optoacoustic Tomography},
  journal = {IEEE Transactions on Instrumentation and Measurement},
  year = {2024},
  volume = {73},
  pages = {1--12},
  doi = {10.1109/tim.2024.3417542}
}
Kostenlos starten →